Journal of Marine Science and Engineering

Article Nearshore Dynamics of Storm Surges and Waves Induced by the 2018 Jebi and Trami Based on the Analysis of Video Footage Recorded on the Coasts of Wakayama,

Yusuke Yamanaka 1,* , Yoshinao Matsuba 1,2 , Yoshimitsu Tajima 1 , Ryotaro Shibata 1, Naohiro Hattori 1, Lianhui Wu 1 and Naoko Okami 1

1 Department of Civil Engineering, The University of Tokyo, Tokyo 113-8656, Japan; [email protected] (Y.M.); [email protected] (Y.T.); [email protected] (R.S.); [email protected] (N.H.); [email protected] (L.W.); [email protected] (N.O.) 2 Research Fellow of Japan Society for the Promotion of Science, Tokyo 102-0083, Japan * Correspondence: [email protected]

 Received: 30 September 2019; Accepted: 11 November 2019; Published: 13 November 2019 

Abstract: In this study, field surveys along the coasts of Wakayama Prefecture, Japan, were first conducted to investigate the coastal damage due to storm surges and storm-induced waves caused by the 2018 Typhoons Jebi and Trami. Special focus was placed on the characteristic behavior of nearshore waves through investigation of observed data, numerical simulations, and image analysis of video footage recorded on the coasts. The survey results indicated that inundation, wave overtopping, and drift debris caused by violent storm-induced waves were the dominant factors causing coastal damage. Results of numerical simulations showed that heights of storm-induced waves were predominantly greater than heights along the entire coast of Wakayama in both typhoons. However, computed gradual alongshore variations in wave and surge heights did not explain locally-concentrated inundation and run-up heights observed along the coasts. These results indicate that complex nearshore hydrodynamics induced by local nearshore bathymetry might have played a significant role in inducing such local wave characteristics and the associated coastal damage. Analysis of video footage recorded during Jebi, for example, clearly showed evidence of amplified infragravity wave components, which could enhance inundation and wave run-up.

Keywords: post-event survey; the 2018 Typhoons Jebi and Trami; gauge data; image analysis; infragravity wave

1. Introduction disasters are one of the major natural disasters in the world, as cyclone-induced winds, rain, waves, and surges cause catastrophic damage in urban areas (e.g., Hurricanes Katrina, Sandy, Irma, and Maria, and ) [1–5]. Intense tropical cyclones, categorized into high-intensity classes (i.e., Categories 4 and 5), are expected to occur more frequently as global warming increases [6,7]. Therefore, individual damages induced by past tropical cyclones in each region should be thoroughly investigated to understand the mechanisms of damage generation and develop better disaster mitigation plans and strategies for the future. Japan has suffered major disasters induced by storms such as Typhoons Vera, Nancy, and Mireille [8–11]. Storm surges and storm-induced waves during a typhoon, as well as winds and rain, largely affect the country’s long coastlines. For example, Typhoon Vera, which crossed over the Kansai

J. Mar. Sci. Eng. 2019, 7, 413; doi:10.3390/jmse7110413 www.mdpi.com/journal/jmse J. Mar. Sci. Eng. 2019, 7, 413 2 of 21 J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 2 of 20 storm-surgeregion, the westerntyphoons part in Japan; of the mainthe maximum island of Japan, surge isheight known was as approximately one of the most 3.6 disastrous m at storm-surge near thetyphoons head of Ise in Japan;Bay [12]. the Countermeasures maximum surge for height storm was surges approximately along the coast 3.6 m in at the Nagoya Kansai nearregion the have head beenof Ise discussed Bay [12]. using Countermeasures virtual typhoon for stormscenarios surges based along on the Typhoons coast in theVera Kansai and regionNancy have [12,13]. been Wakayamadiscussed Prefecture using virtual is one typhoon of the scenariosmost affected based areas on Typhoonsby storm surges Vera and and Nancy storm-induced [12,13]. Wakayama waves in thisPrefecture region because is one ofit has the long most coastlines affected areas facing by the storm open surges Pacific and Ocean, storm-induced extending for waves several in thishundred region kilometers.because it hasIn 2018, long coastlinestwo large facingtyphoons, the openJebi Pacificand Trami, Ocean, passed extending near forthe severalcoast of hundred Wakayama kilometers. and madeIn 2018, landfall two largeon the typhoons, prefecture Jebi and and generated Trami, passed extremely near thedamaging coast of storm Wakayama surges and and made waves landfall on the on coaststhe prefecture (Figure 1). and Many generated structures extremely and houses damaging near the storm coast surges were and inundated waves onand the damaged coasts (Figure during1 ). theseMany typhoons, structures as andreported houses by nearMori theet al. coast [13]. were The inundatedobserved heights and damaged of storm during surges theseat certain typhoons, tide gaugeas reported stations by inMori Wakayama et al. [13 ]were. The the observed highest heights in recorded of storm history surges during at certain either tide Typhoon gauge stations Jebi or in TyphoonWakayama Trami were [14,15]. the highest Consequently, in recorded these history typhoon durings and either the resulting Typhoon damages Jebi or Typhoon may have Trami been [14 the,15 ]. mostConsequently, disastrous theseevents typhoons on the andcoast the of resulting Wakayama damages in the may last have several been thedecades. most disastrousTherefore, events we investigatedon the coast coastal of Wakayama damage inin theWakayama last several induced decades. by these Therefore, two typhoons we investigated based on coastal a post-event damage fieldin Wakayama survey to inducedunderstand by thesethe overall two typhoons nearshor basede hydrodynamic on a post-event characteristics field survey induced to understand by storm the surgesoverall and nearshore storm-induced hydrodynamic waves. Additionally, characteristics in induced this study, by storm data analysis, surges and numerical storm-induced simulations, waves. andAdditionally, image analysis in this of study,the video data footage analysis, recorded numerical on the simulations, coasts were and performed image analysis to estimate of the them video duringfootage the recorded two typhons. on the coasts were performed to estimate them during the two typhons.

Figure 1. Typhoon paths (indicated by lines) and surveyed locations (indicated by colored circles) in Figure 1. Typhoon paths (indicated by lines) and surveyed locations (indicated by colored circles) in Wakayama Prefecture. Red and blue indicate Typhoons Jebi and Trami, respectively. The typhoon paths Wakayama Prefecture. Red and blue indicate Typhoons Jebi and Trami, respectively. The typhoon were based on Japan Meteorological Agency, JMA [16]. The yellow squares indicate tide gauge stations paths were based on Japan Meteorological Agency, JMA [16]. The yellow squares indicate tide gauge at Kushimoto, Shirahama, Gobo, Wakayama, and Kobe from the south, and the yellow triangles indicate stations at Kushimoto, Shirahama, Gobo, Wakayama, and Kobe from the south, and the yellow Nationwide Ocean Wave information for Port and HArborS (NOWPHAS) stations at Shionomisaki, triangles indicate Nationwide Ocean Wave information for Port and HArborS (NOWPHAS) stations Komatsushima, and Kobe from the south. at Shionomisaki, Komatsushima, and Kobe from the south.

J. Mar. Sci. Eng. 2019, 7, 413 3 of 21

2. Post-Event Surveys after the Typhoons

2.1. Damage Caused by Typhoon Jebi passed near the coast of Wakayama during 11:00–13:00 JST (UTC + 9 h) on September 4, 2018, with a central pressure of approximately 950 hPa and the lowest central pressure dropping to 915 hPa [16] (Figure1). This typhoon caused a large disaster because of winds, rain, surges, and waves, not only on the coast of Wakayama but also along the coastal areas of Osaka Bay [13] (Figure1). Japan Society of Civil Engineers (JSCE) immediately organized a joint survey team, the 2018 Typhoon Jebi Coastal Disaster Survey Team, after the typhoon passed; this team surveyed the overall damage mainly on the coasts of Osaka, Hyogo, and Wakayama in the Kansai region [13,17]. Storm surges had a dominant influence near the inner part of Osaka Bay, with maximum inundation heights of + 4.0 and + 9.2 m, including wave run-up effects, from the mean sea level near the head of Tokyo Bay (i.e., Tokyo Peil, TP) [13]. The coasts of Wakayama were also severely damaged by storm surges and storm-induced waves during the typhoon. The authors conducted field surveys during 11 to 13 September and 19 to 21 September 2018, several weeks after the event, to determine the damage generated on the coasts. Horizontal coordinates and elevations of the observed drift-debris, destroyed structures, and witnessed watermarks, due to inundation and wave run-up, were measured by leveling from sea level and using a real-time kinematic global positioning system (RTK-GPS). The authors conducted this survey as part of the JSCE joint survey team, and most of the measured heights can be found in the dataset presented by Mori et al. [13]. The findings of this field survey after Typhoon Jebi are summarized as follows. The authors visited the coasts of Kushimoto, Susami, Shirahama, Tanabe, Minabe, Inami, Gobo, Mihama, Yura, and Arida (Figure1). Tide gauge stations, operated by Japan Meteorological Agency, JMA, at Kushimoto, Shirahama, and Gobo, observed peak surge levels of TP + 1.7 m, TP + 1.6 m, and TP + 3.2 m, respectively, at these locations during Jebi [13,14]. A substantial amount of debris was deposited onshore and covered most parts of the coast for several weeks after the typhoon. Figure2 shows debris deposited on a coastal dike with a crown height of TP + 9.2 m, observed at Tonda Beach in Shirahama; large amounts of debris were also found behind the dike. Inami Town, located in the middle of Wakayama, (Figure1) observed the highest witnessed watermark elevation, TP + 13.2 m, at a fishery harbor [13]. Furthermore, still images and video footages, recorded by a resident, captured large waves 1–2 km north of the port (Figure3), and these videos clearly indicated that waves repeatedly splashed over the dike with a crown height of TP + 7.0 m. These videos captured that water splashed to such a height that it was likely to reach the tops of trees on the dike and that significant water mass with debris flowed with a long-wave component. In Yura Town, located north of Gobo, where the peak surge level was TP + 3.2 m (Figure1), no inundation was observed in the area behind seawalls, while some residents witnessed inundation at a small harbor. According to witnesses, the inundation height and depth were approximately TP + 3.5 and 1.7 m, respectively. Debris also remained on a small coast near the harbor, and the ground level of the debris was nearly TP + 3.5 m. Minatomachi, in Arida City (Figure1), was severely inundated with a ground level of approximately TP + 3 m, and the city was protected by a seawall with a crown height of around TP + 5–7 m [13,18]. Based on photographs taken during and after the typhoon, Yamanaka et al. [18] determined that the maximum inundation depth in the area was 0.6 m and also mentioned that wave overtopping was the primary cause of inundation. J. Mar. Sci. Eng. 2019, 7, 413 4 of 21 J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 4 of 20 J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 4 of 20

Figure 2. Photograph of debris observed on Tonda Beach with a seawall having a crown height of + Figure 2. PhotographPhotograph of of debris debris observed observed on on Tonda Tonda Beac Beachh with with a aseawall seawall having having a acrown crown height height of of+ 9.2 m from the mean sea level near the head of Tokyo Bay (i.e., Tokyo Peil, TP), taken by the authors +9.29.2 m m fromfrom the the mean mean sea sea level level near near the the head head of of To Tokyokyo Bay Bay (i.e., (i.e., Tokyo Tokyo Peil, Peil, TP), TP), taken taken by the authors on September 21, 2018. on September 21, 2018.

Figure 3. Cont.

J. Mar. Sci. Eng. 2019, 7, 413 5 of 21 J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 5 of 20

FigureFigure 3. 3. PhotographsPhotographs recorded recorded by by a a resident resident (a (a) )at at 07:13 07:13 and and (b (b)) at at 12:12 12:12 (i.e., (i.e., before before and and during during TyphoonTyphoon Jebi), on September 4, 2018, at a site in Inami. 2.2. Damage Caused by Typhoon Trami Typhoon Trami, reaching the lowest central pressure of 915 hPa, made landfall on Tanabe in Wakayama Prefecture around 20:00 JST on 30 September 2018, with a central pressure of 950 hPa [16] (Figure1). Typhoon-induced storm surges and waves, similar to those induced by Typhoon Jebi, significantly impacted the coasts of Wakayama, especially the coastal areas from Kushimoto to Tanabe, located on the east side of the typhoon path. Post-disaster survey following Typhoon Trami was conducted by the authors during 2 to 3 October 2018. Following the post-disaster survey after Typhoon Jebi, elevations of the remaining debris, destroyed structures, and watermarks witnessed by residents were measured by the RTK-GPS. Representative damages observed on the coast of Wakayama are presented in detail in this section. The authors visited the coasts of Kushimoto, Susami, Shirahama, Tanabe, Minabe, Inami, Gobo, and Mihama (Figure1). Tide gauge stations at Kushimoto, Shirahama, and Gobo indicated peak surge levels of TP + 2.5 m, TP + 2.0 m, and TP + 3.0 m during Trami, respectively [15]. Similar to Typhoon Jebi, a considerable amount of debris was carried onshore and covered the coasts after Trami. Structures and houses near the coast were severely damaged by storm-induced waves during the typhoon. Figure4 shows a damaged breakwater observed at Suganohama fishery harbor in Kushimoto. In the harbor, debris remained at a ground elevation of TP + 3.6 m, and a resident witnessed waves running up to the same elevation. However, the crown height of the breakwater was TP + 5.1 m, and it collapsed only on the inside slope, as shown in the photograph. There were few witnesses to the high surge and wave elevation during Trami because most of the damage caused by the typhoon was at night; however, important evidence was recorded by a resident at Mirozu station, where a part of a small station building had collapsed at a ground elevation of around TP + 9.3 m. The resident witnessed watermarks in/around the station buildings at 6:30 on 1 October, 10 h after the typhoon’s landfall. This stationFigure is located 4. Photograph on a coastal showing cliff andthe damaged thus waves breakwater might have observed splashed at Suganohama up to the elevation fishery harbor of the in station. The MeraKushimoto, area, taken located by onthe Shimohayaauthors on 2 BayOctober in Tanabe, 2018. on which the typhoon made landfall, suffered damage from waves. Inundation was dominantly caused by overtopping waves during Jebi [13] and was more severe during Trami.A witness stated that the inundation depth during Trami in the area

J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 5 of 20

J. Mar. Sci. Eng. 2019, 7, 413 6 of 21 was approximately 1 m. Furthermore, the inundating flow razed a tree behind the seawall, and part of the tree hit and significantly damaged a nearby house. Therefore, the influence of waves was significant Figure 3. Photographs recorded by a resident (a) at 07:13 and (b) at 12:12 (i.e., before and during in this area. However, these waves were expected to dissipate because of their geometries [13]. Typhoon Jebi), on September 4, 2018, at a site in Inami.

Figure 4. Photograph showing the damaged breakwater observed at Suganohama fishery harbor in Figure 4. Photograph showing the damaged breakwater observed at Suganohama fishery harbor in Kushimoto, taken by the authors on 2 October 2018. Kushimoto, taken by the authors on 2 October 2018. 2.3. Comparison of Inundation Characteristics of Jebi and Trami

Figure5 shows the elevations of the witnessed watermarks, debris, and destroyed structures observed along the coasts of Wakayama from both typhoons. Most of these elevations are equivalent to the run-up height of the inundating flows, primarily induced by overtopping waves. The maximum elevations of these watermarks due to Typhoons Jebi and Trami were TP + 13.2 m at Inami and TP + 10.2 m at Shirahama, respectively. As seen in Figure5 with Figure1, the observed elevations of the watermarks after Typhoon Jebi were relatively higher along the coasts from Shirahama to Mihama, located in the middle of the coast of Wakayama, than those along the northern and southern coasts. Furthermore, the relatively higher elevations were locally concentrated around Shirahama, Inami, and Mihama. The highest observed watermarks caused by Typhoon Trami were in a different location from those caused by Typhoon Jebi. Relatively high watermarks were found both on the east and west sides of the typhoon path. Although survey data of the run-up heights after Trami were not enough to fully capture their alongshore variation, some of the relatively higher run-up heights were observed in identical locations (e.g., Shirahama, Inami, and Mihama after both Jebi and Trami). The recorded peak tide levels during these two typhoons were much lower than the elevations of the watermarks and debris, and this feature clearly indicates the dominant influence of storm-induced waves rather than storm surges themselves along the coast of Wakayama. Additionally, the observed elevations for both typhoons were locally concentrated, and the sites showing high elevations were identical for both typhoons. Therefore, it is important to estimate nearshore wave dynamics to understand the local concentration characteristics of waves and surges. J. Mar. Sci. Eng. 2019, 7, 413 7 of 21 J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 7 of 20

Figure 5. ElevationsElevations of of remaining remaining debris debris and and witnesse witnessedd watermarks watermarks during during Typhoons Typhoons Jebi and Trami [TP m], indicated by red and blue,blue, respectivelyrespectively (Google(Google Earth;Earth; MoriMori etet al.al. [[13]).13]).

3. Numerical Simulation 3. Numerical Simulation To determine the characteristics of storm-induced waves and storm surges around the coast of To determine the characteristics of storm-induced waves and storm surges around the coast of Wakayama during these typhoons, numerical simulations were carried out. Storm-induced wave fields Wakayama during these typhoons, numerical simulations were carried out. Storm-induced wave were simulated based on Wave Watch III (WW3) [19], and storm surges were simulated based on the fields were simulated based on Wave Watch III (WW3) [19], and storm surges were simulated based nonlinear long-wave model with forcing terms due to wind shear stress and an on the nonlinear long-wave model with forcing terms due to wind shear stress and an atmospheric gradient [20]. Bathymetry data of the computation domains for these simulations were obtained from pressure gradient [20]. Bathymetry data of the computation domains for these simulations were the General Bathymetric Chart of the Oceans [21]. The domain of WW3 applied a one-way, nested-grid obtained from the General Bathymetric Chart of the Oceans [21]. The domain of WW3 applied a one- system with resolutions of 30, 12, 3, and 1 arc-min, while the domain for storm surge applied a spherical way, nested-grid system with resolutions of 30, 12, 3, and 1 arc-min, while the domain for storm surge coordinate system with a single grid size of 30 arc-sec. The wind-induced, air-sea drag coefficient applied a spherical coordinate system with a single grid size of 30 arc-sec. The wind-induced, air-sea model of Oey et al. [22] was applied for both computations. WW3 computations were based on a wind drag coefficient model of Oey et al. [22] was applied for both computations. WW3 computations were field obtained from Global Spectral Model - Grid Point Value, GSM-GPV data [23]. The present surge based on a wind field obtained from Global Spectral Model - Grid Point Value, GSM-GPV data [23]. model did not account for the force due to wave radiation stress. Wind and pressure gradient fields The present surge model did not account for the force due to wave radiation stress. Wind and were determined using an empirical typhoon model with the track and central pressure of the typhoon pressure gradient fields were determined using an empirical typhoon model with the track and estimated by JMA [16]. Three different constant values of Manning’s roughness (i.e., 0.010, 0.025, and central pressure of the typhoon estimated by JMA [16]. Three different constant values of Manning’s 0.050 m 1/3s) for representing bottom shear stress were tested. The present model did not account for roughness− (i.e., 0.010, 0.025, and 0.050 m−1/3s) for representing bottom shear stress were tested. The presenttime variations model did in the not astronomical account for time tide, variations and the initial in the still astronomical water level wastide, setand to the TP initial+ 0 m. still water level Figurewas set6 comparesto TP + 0 m. the simulation results with the observed data obtained at NOWPHAS and at tide gaugesFigure 6 (Figure compares1) for the Typhoons simulation Jebi results and Trami.with the While observed some data of the obtained wave dataat NOWPHAS are missing, and the at tidesimulation gauges results (Figure moderately 1) for Typhoons reproduced Jebi theand observed Trami. While time seriessome ofofsignificant the wave wavedata are heights missing, for both the Jebisimulation and Trami. results As moderately the simulated reproduced surge heights the observ are insensitiveed time series to the of roughness significant values wave (Figureheights A1 for), 1/3 bothFigure Jebi6 presents and Trami. the surge As the waveforms simulated simulated surge heig byhts assuming are insensitive a Manning’s to the roughnessroughness ofvalues 0.025 (Figure m − s. A1),In the Figure comparison 6 presents of storm the surge surges, waveforms the observed simulated data show by assuming the difference a Manning’s between roughness the observed of 0.025 tide level and the corresponding astronomical tide so that it can be compared with the simulated surge m−1/3s. In the comparison of storm surges, the observed data show the difference between the observedheights. Totide roughly level and account the corresponding for the influence astronomic of the waveal tide setup, so that Figure it can6c,d be shows compared the simulated with the simulated surge heights. To roughly account for the influence of the wave setup, Figures 6c,d shows the simulated instantaneous surge heights and the ones with a height equal to 5% of the nearshore

J. Mar. Sci. Eng. 2019, 7, 413 8 of 21 instantaneousJ. Mar. Sci. Eng. surge 2019, heights7, x FOR PEER and REVIEW the ones with a height equal to 5% of the nearshore significant 8 of 20 wave heightsignificant simulated wave by WW3.height Assimulated seen in by these WW3. figures, As seen the in simulated these figures, surge the heights simulated reasonably surge heights reproduce the observedreasonably storm reproduce surge height.the observed The simulation storm surge results height. showed The bettersimulation predictions results byshowed considering better the roughlypredictions estimated by waveconsidering setup the (i.e., roughly 5% of estimated nearshore wa waveve setup heights) (i.e., 5% compared of nearshore to the wave tide heights) gauge data, particularlycompared at theto the tide tide gauge gauge stations data, particularly of Kushimoto, at the Shirahama, tide gauge stations and Gobo. of Kushimoto, This feature Shirahama, indicates the significantand Gobo. impact This of feature storm-induced indicates the waves significant on the impact total of increase storm-induced in water waves surface on the elevation total increase along the coastin of water Wakayama. surface elevation along the coast of Wakayama.

Figure 6. Comparisons of observed data and simulation results for Typhoons Jebi and Trami. The observedFigure 6. andComparisons simulated of significant observed data wave and heights simulation are, respectively,results for Typhoons indicated Jebi by and symbols Trami. and Thesolid observed and simulated significant wave heights are, respectively, indicated by symbols and solid lines at NOWPHAS stations for (a) Jebi and (b) Trami; the triangles, squares, and circles, respectively, lines at NOWPHAS stations for (a) Jebi and (b) Trami; the triangles, squares, and circles, respectively, indicate observations at Shionomisaki, Komatsushima, and Kobe; observed surge height, simulated indicate observations at Shionomisaki, Komatsushima, and Kobe; observed surge height, simulated surgesurge height, height, and and simulated simulated surge surge height height with with 55 %% ofof the simulated simulated significant significant wave wave height height at tide at tide gaugegauge stations, stations, respectively, respectively, indicated indicated by by solid solid black,black, solid red, red, and and dashed dashed red red lines lines for ( forc) Jebi (c) Jebiand and (d) Trami.(d) Trami.

J. Mar. Sci. Eng. 2019, 7, 413 9 of 21

J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 9 of 20 Figure7 shows spatial distributions of the simulated maximum values of significant wave heights and stormFigure surge 7 heightsshows spatial during distributions Jebi and Trami. of the si Asmulated seen inmaximum Figure7 a,b,values significant of significant wave wave heights graduallyheights decrease and storm toward surge Kii heights Channel duri andng Jebi Osaka and Bay,Trami. where As seen fetch-limited in Figure 7a, waves b, significant are dominant, wave and peakedheights around gradually the middle decrease coasts toward from Kii Shirahama Channel toand Mihama Osaka Bay, for Jebi where and fetch-limited at the south waves of Kushimoto are dominant, and peaked around the middle coasts from Shirahama to Mihama for Jebi and at the south for Trami. According to the simulation results shown in Figure7c,d, surges in Jebi were locally of Kushimoto for Trami. According to the simulation results shown in Figure 7c, d, surges in Jebi concentratedwere locally around concentrated confined around sea areas confined such sea as Osakaareas such Bay as because Osaka Bay of the because typhoon of the path typhoon and geometry;path on theand other geometry; hand, on insignificant the other hand, concentration insignificant appeared concentration in Osaka appeared Bay in during Osaka TramiBay during because Trami it was locatedbecause to the it west was oflocated the typhoon to the west path. of Thesethe typhoon results path. yielded These the results large contrastyielded the between large contrast Figure 7c,d. The computedbetween Figure maximum 7c and surge 7d. The heights computed on the maximum coast of surge Wakayama heights haveon the a coas peakt of in Wakayama the north forhave Jebi a and aroundpeak Tanabe in the fornorth Trami for Jebi because and around of their Tanabe proximities for Trami to because the typhoon of their paths. proximities Additionally, to the typhoon the model predictspaths. smaller Additionally, storm surges the model to the predicts west ofsmaller Typhoon storm Trami surges path to the than west those of Typhoon to the east, Trami as path evidenced than by the observationsthose to the sucheast, as as evidenced at Wakayama by the located observations to the west such andas at Shirahama Wakayama located located toto thethe eastwest (Figure and 6). Shirahama located to the east (Figure 6). However, local variations in significant wave heights and However, local variations in significant wave heights and surge heights along the coast of Wakayama surge heights along the coast of Wakayama were not noteworthy, while elevations observed in the were not noteworthy, while elevations observed in the field surveys showed substantial variations field surveys showed substantial variations along the coast (Figure 7). Therefore, this feature alongindicates the coast the (Figure importance7). Therefore, of local wave this geometry feature indicates in alongshore the importancevariations. of local wave geometry in alongshore variations.

Figure 7. Distribution of the maximum simulated wave heights during Typhoons Jebi and Trami. Figure 7. Distribution of the maximum simulated wave heights during Typhoons Jebi and Trami. Significant wave height for (a) Jebi and (b) Trami; surge height for (c) Jebi and (d) Trami. Significant wave height for (a) Jebi and (b) Trami; surge height for (c) Jebi and (d) Trami.

J. Mar. Sci. Eng. 2019, 7, 413 10 of 21

4. Analysis of Tide Gauge Records As shown in Figure6c, d, the tide gauge records obtained at Kushimoto, Shirahama, and Gobo show relatively small surge heights during both typhoons, while significant fluctuations in the lower components, other than the wind-wave component, were influential on the coasts. The Power Spectrum Density of fluctuations in the water level recorded at each Tide Gauge (PSDTG) was then estimated based on the Fourier transform, with a sampling data length of 26 and sampling time step of 15 s (corresponding to 16 min window length), to determine if the lower components of the wave amplitude were significant (Figure8). Figure8 indicates that the largest PSDTG peak appears for all components lower than 5.0 10–3 Hz, independent of the locations and typhoons; its peak component for each × site is similar for both typhoons. The figure also shows that all PSDTGs have peaks for components higher than 5.0 10–3 Hz. For example, the PSDTGs of Gobo, Shirahama, and Kushimoto for Jebi and × Trami have peaks around 2.2 10–2, 1.2 10–2, and 1.6 10–2 Hz, respectively. These components × × × are within the range of infragravity wave components (5.0 10–3–4.0 10–2 Hz); therefore, these × × could result from the amplification of waves during nearshore propagation involving reflection and wave breaking. However, the peak components are different for the three sites; this result implied that local bathymetry selectively amplified specific infragravity wave components during propagation independent of storm event. Figure9 shows the estimated wave height corresponding to the spectral components of the infragravity wave ranging from 3.6 10–3 to 3.3 10–2 Hz; it is compared with × × low-pass filtered waveforms obtained from observed waveforms of the surge heights, corresponding to those in Figure6c,d, for visualizing average surge heights over time. As seen in the figure, the peak of the estimated wave height matches with that of the averaged surge height, especially at Gobo, where wave heights during both typhoons reached approximately 2 m. Therefore, these waves may have largely contributed to the transport of debris to the high elevations behind the dike. Additionally, most cases indicate that wave heights were significantly large compared to average surge heights. Consequently, infragravity wave components largely impacted the nearshore water surface fluctuations, especially along the coasts from Kushimoto to Gobo. J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 10 of 20 J. Mar. Sci. Eng. 2019, 7, 413 11 of 21

Figure 8. Power spectrum density of water surface fluctuations observed by tide gauges for (a) Jebi andFigure (b) Trami. 8. Power Eight spectrum power spectrum density of densities water surface are shown fluc fortuations Jebi based observed on water by tide surface gauges fluctuations for (a) Jebi observedand (b)within Trami. time Eight bands power of 9:00–10:00,spectrum densities 10:00–11:00, are shown 11:00–12:00, for Jebi 12:00–13:00, based on water 13:00–14:00, surface 14:00–15:00, fluctuations 15:00–16:00,observed within and 16:00–17:00 time bands and of for9:00–10:00, Trami based 10:00–11:00, on observations 11:00–12:00, within 12:00–13:00, time bands 13:00–14:00, of 16:00–17:00, 14:00– 17:00–18:00,15:00, 15:00–16:00, 18:00–19:00, and 19:00–20:00. 16:00–17:00 20:00–21:00, and for Trami 21:00–22:00, based on 22:00–23:00,observations and within 23:00–24:00. time bands The resultsof 16:00– for17:00, time bands17:00–18:00, 11:00–15:00 18:00–19: and00, 18:00–22:00, 19:00–20:00. when 20:00–21:00, Typhoons 21:00–22:00, Jebi and Trami 22:00–23:00, significantly and 23:00–24:00. impacted the The coasts,results respectively, for time bands are highlighted 11:00–15:00 in redand and 18:00–22:00, blue, respectively. when Typhoons Jebi and Trami significantly impacted the coasts, respectively, are highlighted in red and blue, respectively.

4. Analysis of Tide Gauge Records As shown in Figure 6c, d, the tide gauge records obtained at Kushimoto, Shirahama, and Gobo show relatively small surge heights during both typhoons, while significant fluctuations in the lower components, other than the wind-wave component, were influential on the coasts. The Power Spectrum Density of fluctuations in the water level recorded at each Tide Gauge (PSDTG) was then estimated based on the Fourier transform, with a sampling data length of 26 and sampling time step of 15 s (corresponding to 16 min window length), to determine if the lower components of the wave

J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 11 of 20

amplitude were significant (Figure 8). Figure 8 indicates that the largest PSDTG peak appears for all components lower than 5.0 × 10–3 Hz, independent of the locations and typhoons; its peak component for each site is similar for both typhoons. The figure also shows that all PSDTGs have peaks for components higher than 5.0 × 10–3 Hz. For example, the PSDTGs of Gobo, Shirahama, and Kushimoto for Jebi and Trami have peaks around 2.2 × 10–2, 1.2 × 10–2, and 1.6 × 10–2 Hz, respectively. These components are within the range of infragravity wave components (5.0 × 10–3–4.0 × 10–2 Hz); therefore, these could result from the amplification of waves during nearshore propagation involving reflection and wave breaking. However, the peak components are different for the three sites; this result implied that local bathymetry selectively amplified specific infragravity wave components during propagation independent of storm event. Figure 9 shows the estimated wave height corresponding to the spectral components of the infragravity wave ranging from 3.6 × 10–3 to 3.3 × 10–2 Hz; it is compared with low-pass filtered waveforms obtained from observed waveforms of the surge heights, corresponding to those in Figure 6c,d, for visualizing average surge heights over time. As seen in the figure, the peak of the estimated wave height matches with that of the averaged surge height, especially at Gobo, where wave heights during both typhoons reached approximately 2 m. Therefore, these waves may have largely contributed to the transport of debris to the high elevations behind the dike. Additionally, most cases indicate that wave heights were significantly large compared to average surge heights. Consequently, infragravity wave components largely impacted the nearshore J. Mar. Sci.water Eng. surface2019, 7, fluctuations, 413 especially along the coasts from Kushimoto to Gobo. 12 of 21

Figure 9. Significant wave heights for infragravity wave components (circle) and low-pass filtered waveforms (solid line) from observed waveforms of surge heights with components lower than Figure–4 9. Significant wave heights for infragravity wave components (circle) and low-pass filtered 5.6 10 Hz (30 min) for (a) Jebi and (b) Trami. Significant wave heights, Hm , were calculated × waveforms (solid line) from observed waveforms of surge heights with components0 lower than 5.6 × as Hm0 =–4 4 √m0, where m0 is the zeroth-order moment of power spectrum densities of water level 10 Hz (30 min) for (a) Jebi and (b) Trami. Significant wave heights, 𝐻 , were calculated as 𝐻 = fluctuation; estimation of power spectrum density was based on Fourier transform with a sampling 4𝑚 , where 𝑚 is the zeroth-order moment of power spectrum densities of water level fluctuation; 7 data lengthestimation of 2 ofand power sampling spectrum time density step ofwas 15 based s (corresponding on Fourier transform to 32 min with window a sampling length). data length of 27 and sampling time step of 15 s (corresponding to 32 min window length). 5. Image Analysis of Video Footage Recorded at the Coast The field survey results, numerical analysis of offshore waves and storm surges, and evaluation of tide gauge records clearly indicated that the observed locally-concentrated inundation and wave run-up characteristics were significantly dependent on the complex nearshore hydrodynamics affected by the local nearshore geometry. Similar, locally varying coastal hazards also have been found in other events, and the behavior of infragravity waves could be an important factor in such locally concentrated coastal hazards [24,25]. However, lack of in-situ data from such severe events makes it difficult to understand the detailed features of such complex nearshore hydrodynamics. In recent disasters, several video footages were recorded by local residents, and analysis of these footages could be extremely valuable for enhanced understanding of such complex nearshore hydrodynamic characteristics [26,27]. Because Typhoon Jebi passed near Wakayama in the daytime, several pictures and video footages were recorded by local residents [13,18]. Among the collected video footages, this study focused on two videos that clearly captured the behavior of nearshore waves at different sites (Figure 10) and image analyses were conducted to extract quantitative characteristics of the captured nearshore hydrodynamics.

5.1. Video Footage Recorded at Kushimoto Town The first video was recorded on the coast in front of Kushimoto Marine Park in Kushimoto (33◦28051”N, 135◦44042”E) from approximately 11:00 to 12:00 on 4 September (Figure 10a). The video camera was fixed outside the building. Violent winds slightly vibrated the fixed camera, and drops of rain and waves on the camera affected the quality of some recordings. We carefully investigated the video quality and selected a 12-min continuous footage at approximately 11:40 when the influences were not significant. Figure 11a,b shows still images of the water surface and pier with the steel poles. Because the vertically-installed poles were nearly parallel to the column of the image pixels, line images J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 12 of 20

5. Image Analysis of Video Footage Recorded at the Coast The field survey results, numerical analysis of offshore waves and storm surges, and evaluation of tide gauge records clearly indicated that the observed locally-concentrated inundation and wave run-up characteristics were significantly dependent on the complex nearshore hydrodynamics affected by the local nearshore geometry. Similar, locally varying coastal hazards also have been found in other events, and the behavior of infragravity waves could be an important factor in such locally concentrated coastal hazards [24,25]. However, lack of in-situ data from such severe events makes it difficult to understand the detailed features of such complex nearshore hydrodynamics. In recent disasters, several video footages were recorded by local residents, and analysis of these footages could be extremely valuable for enhanced understanding of such complex nearshore hydrodynamic characteristics [26,27]. Because Typhoon Jebi passed near Wakayama in the daytime, several pictures and video footages were recorded by local residents [13,18]. Among the collected video footages, this study focused on two videos that clearly captured the behavior of nearshore waves at different sites (Figure 10) and image analyses were conducted to extract quantitative characteristics of the captured nearshore hydrodynamics.

5.1. Video Footage Recorded at Kushimoto Town The first video was recorded on the coast in front of Kushimoto Marine Park in Kushimoto (33°28′51″N, 135°44′42″E) from approximately 11:00 to 12:00 on 4 September (Figure 10a). The video camera was fixed outside the building. Violent winds slightly vibrated the fixed camera, and drops of rain and waves on the camera affected the quality of some recordings. We carefully investigated J. Mar. Sci.the Eng. video2019 ,quality7, 413 and selected a 12-min continuous footage at approximately 11:40 when the 13 of 21 influences were not significant. Figure 11a,b shows still images of the water surface and pier with the steel poles. Because the vertically-installed poles were nearly parallel to the column of the image along thepixels, column line images at a pole along location the column were at a extracted pole location from were succeeding extracted from still succeeding images of still the images video, of and the extractedthe line video, images and the were extracted horizontally line images placed were horizont in the orderally placed of the in the recorded order of time the recorded to create time a time to stack image ofcreate the fluctuatinga time stack image water of surface the fluctuating levels (Figurewater surface 12a). levels Figure (Figure 11a,b 12a). indicates Figure blackish11a,b indicates and whitish colors ofblackish the pier and and whitish water colors surface of the pier in the and recorded water surface images; in the thus,recorded water images; surface thus, elevationwater surface along the elevation along the sidewall of the pier and pole can be easily detected as the boundary between the sidewall of the pier and pole can be easily detected as the boundary between the blackish and whitish blackish and whitish pixels (Figure 12a). Time, t, refers to the elapsed time from the first image of the pixels (Figureselected 12video;a). Time,the heightt, refers and elevation to the elapsed in the time time stack from image the were first based image on the of thepole’s selected height and video; the height andelevation. elevation in the time stack image were based on the pole’s height and elevation.

Figure 10. Overview of the geometry based on the Geospatial Information Authority of Japan [28] and Figure 10. Overview of the geometry based on the Geospatial Information Authority of Japan [28] locationsJ. Mar. video Sci. Eng. site; 2019 (a, 7), Kushimotox FOR PEER REVIEW (Kushimoto Marine Park) and (b) Obiki area in Yura Town. 13 of The 20 red and locations video site; (a) Kushimoto (Kushimoto Marine Park) and (b) Obiki area in Yura Town. circles and red vectors indicate video locations and video angles, respectively. The red circles and red vectors indicate video locations and video angles, respectively.

Figure 11. FigureStill images 11. Still images from thefrom video the video footage footage when when the the water water level level (a) decreased (a) decreased and (b) andincreased, (b) increased, respectively,respectively, taken at taken Kushimoto at Kushimoto Marine Marine Park Park Ltd. Ltd aroundaround 11:45 11:45 on on4 September 4 September 2018. The 2018. red lines The in red lines in the imagesthe indicate images indicate the pixels the pixels for creating for creating the the time time stackstack image. image.

J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 13 of 20

Figure 11. Still images from the video footage when the water level (a) decreased and (b) increased, respectively, taken at Kushimoto Marine Park Ltd around 11:45 on 4 September 2018. The red lines in J. Mar. Sci.the Eng.images2019 indicate, 7, 413 the pixels for creating the time stack image. 14 of 21

J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 14 of 20

Figure 12. (a) Time stack image for estimating water-level fluctuations indicated by red lines; Figure(b) estimated 12. (a) water-levelTime stack fluctuationsimage for estimating based on Figurewater-level 12a with fluctuations synthesized indicated waveforms by red comprising lines; (b) wind-waveestimated water-level and infragravity fluctuations wave components, based on Fi respectively,gure 12a with indicated synthesized in red, blue, waveforms and black; comprising (c) Power wind-waveSpectrum Densities and infragravity based on wave Tide components, Gauge data andresp water-levelectively, indicated fluctuations in red, extracted blue, and from black; Video (c) PowerFootage, Spectrum PSDTG (black)Densities and based PSDVF on (red). Tide Gauge data and water-level fluctuations extracted from Video Footage, PSDTG (black) and PSDVF (red). Figure 12b indicates the extracted water level fluctuations from the time stack image (Figure 12a) and its frequency components from 4.0 10–2 to 5.0 10–1 Hz and lower than 4.0 10–2 Hz, respectively, Figure 12b indicates the extracted ×water level flu× ctuations from the time stack× image (Figure 12a) andbased its onfrequency spectral components analysis. Each from component 4.0 × 10–2 to respectively 5.0 × 10–1 Hz represents and lower the than wind 4.0 wave× 10–2 Hz, and respectively, infragravity wavebased components.on spectral analysis. Figure 12Eachc compares component the respectively Power Spectrum represents Density the ofwind water wave surface and fluctuationsinfragravity waveextracted components. from the VideoFigure Footage 12c comp (PSDVF)ares the and Power PSDTG Spectrum of Kushimoto, Density locatedof water 3 surface km east fluctuations of the park, extractedfrom 12:00 from to 13:00,the Video as displayed Footage (PSDVF) in Figure and8. A PSDTG Fourier of transformKushimoto, for located estimating 3 km theeast PSDVF of the park, was 11 fromconducted 12:00 withto 13:00, a sampling as displayed data lengthin Figure of 28. Aand Fourier sampling transform time step for ofestimating 0.1 s. First, the asPSDVF shown was in conductedFigure 12b, with the amplitudea sampling of data the length low-frequency of 211 and components sampling time (i.e., step infragravity of 0.1 s. First, wave as components) shown in Figure was 12b,approximately the amplitude 1 m; thisof the amplitude low-frequency was as high components as that of the(i.e., wind-wave infragravity components wave components) and thus had was a approximatelysignificant impact 1 m; on this inundation amplitude or wavewas as run-up high as along that theof the coast wind-wave of Kushimoto. components Furthermore, and thus as shown had a –3 –3 –2 significantin Figure12 impactc, three on peaks inundation appear ator 2.1wave 10run-up, 5.0 along10 the, and coast 1.4 of 10Kushimoto.Hz in PSDTG. Furthermore,Endoh as et shown al. [29] × × × ininvestigated Figure 12c, athree resonant peaks oscillationappear at 2.1 excited × 10–3, in5.0 Kushimoto × 10–3, and 1.4 Bay × and10–2 Hz determined in PSDTG. the Endoh predominant et al. [29] investigatedcomponent (i.e., a resonant the first modeoscillation bay-scale excited resonance in Kushimoto period) to Bay be 7.75and determ10–4 Hz.ined Endoh the etpredominant al. [29] also × component (i.e., the first mode bay-scale resonance period) to be 7.75 × 10–4 Hz. Endoh et al. [29] also showed that the frequency components ranging from 1.0 × 10–3 to 2.0 × 10–3 Hz, corresponding to the peak at 2.1 × 10–3 observed in PSDTG, were amplified. The other peak frequencies of PSDTG (5.0 × 10–3 and 1.4 × 10–2 Hz) were higher than the first peak component and this may be due to the resonance of infragravity waves in local bathymetry. On the other hand, PSDVF showed two peaks at 1.0 × 10–2 and 5.0 × 10–2–6.0 × 10–2 Hz (Figure 12c). The latter frequency was equivalent to the significant wave period of 14–15 s simulated by WW3 and thus should represent wind wave components. The former frequency was similar to the frequency of one of the PSDTG peaks, at 1.4 × 10–2 Hz. However, PSDTG at 1.4 × 10–2 Hz was much smaller than the one for PSDVF at 1.0 × 10–2 Hz. Tide gauge data were recorded with time intervals of 15 s, and the high-frequency components of water-level fluctuations were physically dissipated by a conduit connecting the tide gauge and open sea. Namegaya et al. [30] also indicated that tide gauge data tend to underestimate water-level fluctuations for relatively shorter frequency components, including tsunami waves. Considering the estimated results into account, the infragravity wave components, at approximately 1.4 × 10–2 and 5.0 × 10–3 Hz, were enhanced by local geometry of Kushimoto and might dominate nearshore wave dynamics there.

J. Mar. Sci. Eng. 2019, 7, 413 15 of 21 showed that the frequency components ranging from 1.0 10–3 to 2.0 10–3 Hz, corresponding to the × × peak at 2.1 10–3 observed in PSDTG, were amplified. The other peak frequencies of PSDTG (5.0 10–3 × × and 1.4 10–2 Hz) were higher than the first peak component and this may be due to the resonance of × infragravity waves in local bathymetry. On the other hand, PSDVF showed two peaks at 1.0 10–2 × and 5.0 10–2–6.0 10–2 Hz (Figure 12c). The latter frequency was equivalent to the significant wave × × period of 14–15 s simulated by WW3 and thus should represent wind wave components. The former frequency was similar to the frequency of one of the PSDTG peaks, at 1.4 10–2 Hz. However, PSDTG × at 1.4 10–2 Hz was much smaller than the one for PSDVF at 1.0 10–2 Hz. Tide gauge data were × × recorded with time intervals of 15 s, and the high-frequency components of water-level fluctuations were physically dissipated by a conduit connecting the tide gauge and open sea. Namegaya et al. [30] also indicated that tide gauge data tend to underestimate water-level fluctuations for relatively shorter frequency components, including tsunami waves. Considering the estimated results into account, the infragravity wave components, at approximately 1.4 10–2 and 5.0 10–3 Hz, were enhanced by × × local geometry of Kushimoto and might dominate nearshore wave dynamics there.

5.2. Video Footage Recorded at Yura Town The second image-based analysis was conducted for the video footage recorded at a site behind the coast in the Obiki area, Yura (33◦58004”N, 135◦05014”E), at approximately 13:00 on 4 September (Figure 10b). The 30 s video footage captured a bore-like wave that continuously propagated in the upward direction along the 3 m wide rectangular open channel, connected to the open sea (Figures 10a and 13a–c). Time, t, refers to the elapsed time from the video’s first images. Based on the video footage and survey results, the maximum height of this bore wave was estimated to be approximately 1.8 m. To quantitatively estimate the wave characteristics, the video was analyzed using the rectification technique proposed by Holland et al. [31]. However, a preprocessing step was necessary to analyze the video footage, because the method described by Holland et al. [31] requires fixed ground control points, while the video had unfixed image frames. Therefore, we applied a template-matching technique based on the zero-mean normalized cross-correlation (ZNCC) to RGB brightness in each image to track the control points in the moving frames [32,33]. A small confined area, comprising the number of vertical and horizontal pixels, M and N, respectively, was first determined in a still image (the first image) from the video footage as a template, and a ground control point was reflected in the template. An area with the same dimensions, where RGB distribution was the most similar to the template’s RGB, in the next image, i.e., an image recorded at a certain sec after the first image (the second image), was specified based on ZNCC, and then, the location of the control point in the second image was also specified. Next, the distribution of RGB brightness in the template was replaced with the distribution of that in the specified area in the second image; the template locations of a still image after the second image (the third image) were determined based on ZNCC and the replaced RGB brightness. Therefore, the control points in the moving frames could be tracked by repeating this process until the end of the video. The geometric parameters for all frames were then estimated based on the estimated control points. J. Mar. Sci. Eng. Eng. 20192019,, 77,, x 413 FOR PEER REVIEW 1516 of of 2021

Figure 13. Still images from the recorded video at (a) t = 12.9 s, (b) 18.8 s, and (c) t = 26.9 s, and (d–f) top-view images obtained by rectifying the images in (a–c), respectively, with the control line. Figure 13. Still images from the recorded video at (a) t = 12.9 s, (b) 18.8 s, and (c) t = 26.9 s, and (d–f) top-viewFigure 13 imagesd–f shows obtained top-view by rectifying images the for images Figure in 13(a–a–c,c), respectively, respectively, with obtained the control by line. converting the pixel coordinates into Cartesian XY coordinates using the estimated geometric parameters. As seen in Figure 13, the camera changed to completely different angles very quickly to capture the wave front and wave flows; such situations make tracking same control points difficult. Therefore, the video was 5.2. Video Footage Recorded at Yura Town divided into three sections with similar camera angles, and different control points were tracked in eachThe section second (Figure image-based 13a–c). Figure analysis 13 showswas conducted that the frontfor the face video of the footage bore-like recorded wave at in a thesite channel behind theappears coast whitein theowing Obiki toarea, turbulence. Yura (33°58 Therefore,′04″N, 135°05 all the′14″ imageE), at approximately frames were, rectified 13:00 on into 4 September top-view (Figureimages and10b). a The control 30 s line, video indicated footage by captured the red linesa bore in-like Figure wave 13 d–f,that wascontinuously determined; propagated RGB brightness in the upwardon the line direction was extracted along the to 3 visualize m wide therectangular bore-wave open propagation channel, connected in time and to the space, open as sea a time (Figures stack 10aimage and (Figure 13a–c). 14 Time,). In Figure t, refers 14 ,to the the horizontal elapsed time axis indicatesfrom the thevideo’s distance first alongimages. the Based channel, on lthe, positive video footagein the landward and survey direction, results, the and maximum the vertical height axis of indicates this bore time, wave positive was estimated in the downward to be approximately direction. Therefore,1.8 m. To thequantitatively right-downward estimate line ofthe the wave bore frontcharacteristics, demonstrates the that video the bore-likewas analyzed wave propagatedusing the rectificationlandward along technique the channel, proposed and by the Holland propagating et al. [31]. speed However, of this a wavepreprocessing was approximately step was necessary 4.2 m/s. to analyze the video footage, because the method described by Holland et al. [31] requires fixed ground control points, while the video had unfixed image frames. Therefore, we applied a template-

J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 16 of 20

matching technique based on the zero-mean normalized cross-correlation (ZNCC) to RGB brightness in each image to track the control points in the moving frames [32,33]. A small confined area, comprising the number of vertical and horizontal pixels, M and N, respectively, was first determined in a still image (the first image) from the video footage as a template, and a ground control point was reflected in the template. An area with the same dimensions, where RGB distribution was the most J. Mar.similar Sci. Eng. to 2019the ,template’s7, 413 RGB, in the next image, i.e., an image recorded at a certain sec after17 the of 21 first image (the second image), was specified based on ZNCC, and then, the location of the control point in the second image was also specified. Next, the distribution of RGB brightness in the template was Furthermore, with l = 0–20 m and t= 24–27 s, the water surface was still moving landward. Based replaced with the distribution of that in the specified area in the second image; the template locations on this figure, the bore-like wave continuously progressed landward for more than 15 s. As seen in of a still image after the second image (the third image) were determined based on ZNCC and the Figure 10b, the channel curved, and the along-channel distance from the sea to the video-recording site replaced RGB brightness. Therefore, the control points in the moving frames could be tracked by was more than 150 m. The observed 1.8 m-high bore-like wave propagating more than 150 m along repeating this process until the end of the video. The geometric parameters for all frames were then the open channel also indicates that significantly high infragravity wave components were generated estimated based on the estimated control points. around the coast of Yura.

Figure 14. Time stack image along the control line, as indicated in Figure 13d–f. Figure 14. Time stack image along the control line, as indicated in Figure 13d–f.

J. Mar. Sci. Eng. 2019, 7, 413 18 of 21

6. Discussion and Conclusions We investigated damages along the coast of Wakayama, induced by storm surges and storm-induced waves during Typhoons Jebi and Trami, and estimated how waves appeared near the shore and caused damage. Damage, such as inundation and destruction of structures, during these typhoons was widely observed along the coast. Relatively higher elevations for watermarks and debris were found after both Jebi and Trami, especially at Shirahama, Inami, and Mihama. As indicated by simulation results showing gradual alongshore variation in wave heights and surges during these typhoons, large waves should have affected these coasts without significant dissipation. According to Mori et al. [13], dissipating waves affected the coast of Mera, facing the northwest and protected by small islands, and the waves were; therefore, expected to cause insignificant damage, while severe inundation occurred in the area with Jebi and Trami. As mentioned in Section4, waves with components lower than 5.0 10–3 Hz had the largest amplitude. A resonant oscillation in Shimohaya × Bay, where the Mera area is located, with a period of 5 min or less (higher than 3.3 10–3 Hz) could × be enhanced in a storm-induced wave condition [34]. Therefore, the bay-scale resonance, which developed under storm conditions, also may have been responsible for inundation. Furthermore, large infragravity waves, which developed through coastal resonance, were observed on the coasts of Kushimoto, Shirahama, Gobo, and Yura during Typhoons Jebi and Trami, by tide gauges and in video footages. As mentioned in Section2, the long-wave component, which might correspond to the infragravity wave component, was also qualitatively observed in the video footage recorded in Inami during Typhoon Jebi. Infragravity waves are one of the essential factors in compounding coastal damage. These waves cause significant impacts near the shore [35] and could dominate run-up heights in a swash zone [36]. Video footage recorded during Typhoon Haiyan, for example, showed an infragravity wave that developed on a coral reef, severely damaging the coast in Philippines [26]; these infragravity waves were one of the major factors influencing the coastal hydrodynamics as well as coastal damage [37,38]. These waves may have been largely responsible for coastal damages and elevations of the watermarks observed in Wakayama during these typhoons. Therefore, considering the influence of these waves is necessary for future disaster mitigation planning.

Author Contributions: Formal analysis, Y.Y., Y.M. and N.H.; investigation, Y.Y., Y.M., Y.T., R.S., L.W. and N.O.; writing—original draft, Y.Y.; writing—review and editing, Y.T. Funding: Part of this study was supported by the JSPS KAKENHI Grant Number JP18J21135 and was conducted as a research activity of “Enhancement of National Resilience against Natural Disasters”, a cross-ministerial Strategic Innovation Promotion Program (SIP), under the supervision of the NIED. The program was supported by the Council for Science, Technology, and Innovation (CSTI). Acknowledgments: We thank the editor and two anonymous reviewers for providing helpful comments to improve our manuscript. We appreciate the residents, Susami Tourism Association, and Kushimoto Marine Park Ltd. who provided us with photographs and videos taken during/after the typhoons, and other important information. The GSM-GPV data were originally provided by the JMA. To access these data, the archiving system of Oki/Kanae-Lab and Kitsuregawa-Lab, IIS, Univ. of Tokyo was used. Data is accessible at http://apps.diasjp.net/gpv/. The DIAS dataset is archived and provided under the framework of the Data Integration and Analysis System (DIAS) funded by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT). Several figures presented in this paper were printed using the Generic Mapping Tools (GMT) developed by Wessel et al. [39]. Conflicts of Interest: The authors declare no conflicts of interest. J. Mar. Sci. Eng. 2019, 7, x FOR PEER REVIEW 18 of 20

Author Contributions: Formal analysis, Y.Y., Y.M. and N.H.; investigation, Y.Y., Y.M., Y.T., R.S., L.W. and N.O.; writing—original draft, Y.Y.; writing—review and editing, Y.T..

Funding: Part of this study was supported by the JSPS KAKENHI Grant Number JP18J21135 and was conducted as a research activity of "Enhancement of National Resilience against Natural Disasters", a cross-ministerial Strategic Innovation Promotion Program (SIP), under the supervision of the NIED. The program was supported by the Council for Science, Technology, and Innovation (CSTI).

Acknowledgments: We thank the editor and two anonymous reviewers for providing helpful comments to improve our manuscript. We appreciate the residents, Susami Tourism Association, and Kushimoto Marine Park Ltd. who provided us with photographs and videos taken during/after the typhoons, and other important information. The GSM-GPV data were originally provided by the JMA. To access these data, the archiving system of Oki/Kanae-Lab and Kitsuregawa-Lab, IIS, Univ. of Tokyo was used. Data is accessible at http://apps.diasjp.net/gpv/. The DIAS dataset is archived and provided under the framework of the Data Integration and Analysis System (DIAS) funded by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT). Several figures presented in this paper were printed using the Generic Mapping Tools (GMT) developed by Wessel et al. [39]. J. Mar. Sci. Eng. 2019, 7, 413 19 of 21 Conflicts of Interest: The authors declare no conflicts of interest.

AppendixAppendix AA

Figure A1. Comparisons of simulation results for storm surges in (a) Jebi and (b) Trami. Solid red and 1/3 blueFigure lines A1. indicate Comparisons waveforms of simulation simulated results with afor Manning’s storm surges roughness in (a) Jebi of 0.025and (b m) −Trami.s, corresponding Solid red and toblue waveforms lines indicate displayed waveforms in Figure simulate6. Dottedd with black a Manning’s lines and roughness dashed black of 0.025 lines m indicate−1/3s, corresponding simulated 1/3 waveformsto waveforms with displayed the Manning’s in Figure roughness 6. Dotted of 0.010 black and lines 0.050 and m dashed− s, respectively. black lines indicate simulated waveforms with the Manning’s roughness of 0.010 and 0.050 m−1/3s, respectively. References

1. Cox, D.; Arikawa, T.; Barbosa, A.; Guannel, G.; Inazu, D.; Kennedy, A.; Li, Y.; Mori, N.; Perry, K.; Prevatt, D.; et al. Hurricanes Irma and Maria post-event survey in US Virgin Islands. Coast. Eng. J. 2019, 61, 121–134. [CrossRef] 2. Tajima, Y.; Yasuda, T.; Pacheco, B.M.; Cruz, E.C.; Kawasaki, K.; Nobuoka, H.; Miyamoto, M.; Asano, Y.; Arikawa, T.; Ortigas, N.M.; et al. Initial report of jsce-pice joint survey on the storm surge disaster caused by typhoon haiyan. Coast. Eng. J. 2014, 56, 1450006. [CrossRef] 3. Mori, N.; Kato, M.; Kim, S.; Mase, H.; Shibutani, Y.; Takemi, T.; Tsuboki, K.; Yasuda, T. Local amplification of storm surge by Super Typhoon Haiyan in Leyte Gulf. Geophys. Res. Lett. 2014, 41, 5106–5113. [CrossRef] [PubMed] 4. Sopkin, K.L.; Stockdon, H.F.; Doran, K.S.; Plant, N.G.; Morgan, K.L.M.; Guy, K.K.; Smith, K.E.L. Hurricane Sandy: Observations and Analysis of Coastal Change; US Geological Survey: Reston, VA, USA, 2014. 5. Fritz, H.M.; Blount, C.; Sokoloski, R.; Singleton, J.; Fuggle, A.; McAdoo, B.G.; Moore, A.; Grass, C.; Tate, B. storm surge distribution and field observations on the Mississippi Barrier Islands. Estuarine, Coast. Shelf Sci. 2007, 74, 12–20. [CrossRef] J. Mar. Sci. Eng. 2019, 7, 413 20 of 21

6. Knutson, T.R.; McBride, J.L.; Chan, J.C.L.; Emanuel, K.; Holland, G.; Landsea, C.; Held, I.; Kossin, J.P.; Srivastava, A.K.; Sugi, M. Tropical cyclones and climate change. Nat. Geosci. 2010, 3, 157–163. [CrossRef] 7. Webster, P.J.; Holland, G.J.; Curry, J.A.; Chang, H.-R. Changes in Tropical Cyclone Number, Duration, and Intensity in a Warming Environment. Science 2005, 309, 1844–1846. [CrossRef] 8. Bruneau, N.; Griser, J.; Loridan, T.; Bellone, E.; Khare, S. The impact of extra-tropical transitioning on storm surge and waves in catastrophe risk modeling: Application to the Japanese coastline. Nat. Hazards 2017, 85, 649–667. [CrossRef] 9. Karaca, E.; Aslani, H. Review of two Japan Typhoon catastrophe models for commercial and industrial properties. Nat. Hazards 2016, 83, 19–40. [CrossRef] 10. Kim, S.; Matsumi, Y.; Yasuda, T.; Mase, H. Storm surges along the Tottori coasts following a typhoon. Ocean. Eng. 2014, 91, 133–145. [CrossRef] 11. Kawabata, T.; Kunii, M.; Bessho, K.; Kanazawa, T. Reanalysis and reforecast of Typhoon Vera (1959) using a mesoscale four-dimensional variational assimilation system. J. Meteorol. Soc. Jpn. 2012, 90, 467–491. [CrossRef] 12. Arakawa, H. Typhoon Vera and its Induced Storm Tide. Weatherwise 1960, 13, 150–152. [CrossRef] 13. Mori, N.; Yasuda, T.; Arikawa, T.; Kataoka, T.; Nakajo, S.; Suzuki, K.; Yamanaka, Y.; Webb, A. 2018 Typhoon Jebi post-event survey of coastal damage in the Kansai region, Japan. Coast. Eng. J. 2019, 61, 278–294. [CrossRef] 14. Japan Meteorological Agency. Strong Wind and Storm Surge Caused by Typhoon No. 21 in 2018, (Written in Japanese); Technical Report; Japan Meteorological Agency: Tokyo, Japan, 2018; Available online: http: //www.data.jma.go.jp/obd/stats/data/bosai/report/2018/20180911/20180911.html (accessed on 18 April 2019). 15. Japan Meteorological Agency. Strong Wind and Storm Surge Caused by Typhoon No. 24 in 2018 (Written in Japanese); Technical Report; Japan Meteorological Agency: Tokyo, Japan, 2018; Available online: http: //www.data.jma.go.jp/obd/stats/data/bosai/report/2018/20181011/20181011.html (accessed on 18 April 2019). 16. Japan Meteorological Agency. Track Data of Typhoons in 2018, (in Japanese); Technical Report; Japan Meteorological Agency: Tokyo, Japan, 2018; Available online: https://www.data.jma.go.jp/fcd/yoho/typhoon/ route_map/bstv2018.html (accessed on 6 December 2018). 17. Takabatake, T.; Mäll, M.; Esteban, M.; Nakamura, R.; Kyaw, T.O.; Ishii, H.; Valdez, J.J.; Nishida, Y.; Noya, F.; Shibayama, T. Field Survey of 2018 Typhoon Jebi in Japan: Lessons for Disaster Risk Management. Geosciences 2018, 8, 412. [CrossRef] 18. Yamanaka, Y.; Shibata, R.; Tajima, Y.; Okami, N. Inundation characteristics in Arida City due to overtopping waves induced by 2018 Typhoon Jebi. Proceedings of International Conference on Asian and Pacific Coasts 2019, Hanoi, Vietnam, 25–28 September 2019; pp. 199–206. 19. WW3DG (The WAVEWATCH III Development Group). User Manual and System Documentation of Wavewatch III Version 5.16; Tech. Note 329; NOAA/NWS/NCEP/MMAB: College Park, MD, USA, 2016. 20. Tajima, Y.; Gunasekara, K.H.; Shimozono, T.; Cruz, E.C. Study on Locally Varying Inundation Characteristics Induced by Super Typhoon Haiyan. Part 1: Dynamic Behavior of Storm Surge and Waves Around San Pedro Bay. Coast. Eng. J. 2016, 58, 1640002. [CrossRef] 21. Weatherall, P.; Marks, K.M.; Jakobsson, M.; Schmitt, T.; Tani, S.; Arndt, J.E.; Rovere, M.; Chayes, D.; Ferrini, V.; Wigley, R. A new digital bathymetric model of the world’s oceans. Earth Space Sci. 2015, 2, 331–345. [CrossRef] 22. Ezer, T.; Oey, L.-Y.; Wang, D.-P.; Fan, S.-J.; Yin, X.-Q.; Oey, L.; Wang, D.; Fan, S.; Yin, X. Loop Current warming by Hurricane Wilma. Geophys. Res. Lett. 2006, 33, 08613. [CrossRef] 23. Ikoma, E. GPV Data Archive [Data set]. Data Integration and Analysis System (DIAS). 2010. Available online: https://doi.org/10.20783/DIAS.161 (accessed on 24 April 2019). 24. Tajima, Y.; Shimozono, T.; Gunasekara, K.H.; Cruz, E.C. Study on Locally Varying Inundation Characteristics Induced by Super Typhoon Haiyan. Part 2: Deformation of Storm Waves on the Beach with Fringing Reef Along the East Coast of Eastern Samar. Coast. Eng. J. 2016, 58, 1640003. [CrossRef] 25. Tajima, Y.; Sato, S. Numerical investigations of locally concentrated damages on Seisho coast due to typhoon T0709. In Proceeding of the 5th International Conferences on Asian Pacific Coasts (APAC 2009), Singapore, 13–16 October 2009. 26. Roeber, V.; Bricker, J.D. Destructive tsunami-like wave generated by surf beat over a coral reef during Typhoon Haiyan. Nat. Commun. 2015, 6, 7854. [CrossRef] J. Mar. Sci. Eng. 2019, 7, 413 21 of 21

27. Yamanaka, Y.; Tajima, Y.; Sato, S.; Liu, H. Dynamic behaviors of the 2011 Tohoku Tsunami in Ryori Bay. In Proceedings of the 7th International Conference on Coastal Dynamics, Arcachon, France, 23–26 June 2013; pp. 1937–1946. 28. Geospatial Information Authority of Japan. Basic Map Information. 2019. Available online: http://www.gsi.go. jp/kiban/ (accessed on 30 June 2019). 29. Endoh, T.; Inazu, D.; Waseda, T.; Hibiya, T. A parameter quantifying radiation damping of bay oscillations excited by incident tsunamis. Cont. Shelf Res. 2018, 157, 10–19. [CrossRef] 30. Namegaya, Y.; Tanioka, Y.; Abe, K.; Satake, K.; Hirata, K.; Okada, M.; Gusman, A.R. In situ Measurements of Tide Gauge Response and Corrections of Tsunami Waveforms from the Niigataken Chuetsu-oki Earthquake in 2007. Pure Appl. Geophys. 2009, 166, 97–116. [CrossRef] 31. Holland, K.; Holman, R.; Lippmann, T.; Stanley, J.; Plant, N. Practical use of video imagery in nearshore oceanographic field studies. IEEE J. Ocean. Eng. 1997, 22, 81–92. [CrossRef] 32. Di Stefano, L.; Mattoccia, S.; Tombari, F. ZNCC-based template matching using bounded partial correlation. Pattern Recognit. Lett. 2005, 26, 2129–2134. [CrossRef] 33. Briechle, K.; Hanebeck, U.D. Template matching using fast normalized cross correlation. In Proceedings of the Optical Pattern Recognition XII, Orlando, FL, USA, 19 April 2001. 34. Yoshioka, H.; Takayama, T.; Serizawa, S. Observation and computation of resonant long period oscillation excited by storm wind. In Proceedings of the Coastal Engineering Conference, Lisbon, Portugal, 19–24 September 2004; pp. 1290–1298. 35. Bertin, X.; De Bakker, A.; Van Dongeren, A.; Coco, G.; André, G.; Ardhuin, F.; Bonneton, P.; Bouchette, F.; Castelle, B.; Crawford, W.C.; et al. Infragravity waves: From driving mechanisms to impacts. Earth Sci. Rev. 2018, 177, 774–799. [CrossRef] 36. Sénéchal, N.; Coco, G.; Bryan, K.R.; Holman, R.A. Wave runup during extreme storm conditions. J. Geophys. Res. Space Phys. 2011, 116.[CrossRef] 37. Tajima, Y. Case study on impact of storm waves on inundation characteristics. Procedia IUTAM 2017, 25, 36–43. [CrossRef] 38. Shimozono, T.; Tajima, Y.; Kennedy, A.B.; Nobuoka, H.; Sasaki, J.; Sato, S. Combined infragravity wave and sea-swell runup over fringing reefs by super typhoon Haiyan. J. Geophys. Res. Oceans 2015, 120, 4463–4486. [CrossRef] 39. Wessel, P.; Smith, W.H.F.; Scharroo, R.; Luis, J.; Wobbe, F. Generic Mapping Tools: Improved Version Released. Eos Trans. AGU 2013, 94, 409–410. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).